What it means to pre-register a backtest
Pre-registration comes from clinical and social science research, where it was introduced to stop researchers from choosing the analysis after seeing which one gave the most interesting answer. The same discipline applies to trading research.
What gets written down
- The hypothesis. A specific claim, such as "stocks with the strongest returns over the past several months continue to outperform over the next month."
- The rules. Universe, signal, entry, exit, position sizing and trading costs.
- The parameters to be tried. A fixed list, not an open-ended search.
- The measure of success. The statistic and the threshold, chosen before results exist.
- The test plan. Which periods are used for development, which for confirmation, and which is held back for a single final run.
Why it helps
Without a record made in advance, it is easy to change the question until the answer looks good: a different lookback, a different universe, a different start date. Each change feels reasonable. Together they amount to overfitting, and nobody, including the researcher, can tell afterwards.
What it does not do
Pre-registration does not make a strategy work. It makes the evidence about whether it works more trustworthy, including when the answer is no.
Where we use it
Our methodology records the hypotheses and the test plan first, and the research log reports against them.
This article is educational and is not investment advice.